Unveiling Neural Networks for Personalized Diet Recommendations

• Main Author: Cunha, Carlos
• Publication Date: 2024
• Other Authors: Rebelo, João, P. Duarte, Rui
• Format: Article
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10400.19/8597
• topic: Personalized Nutrition; Error Prediction; Machine Learning

The growing prevalence of poor nutrition is a major public health concern, as it fuels the rise of various diseases. Obesity, a silent and rapidly growing threat linked to unhealthy eating, is a prime example. Despite the abundance of information on diets and recipes, finding a personalized approach to healthy eating can be a challenge. Recommendation systems can filter from a food logging dataset the information that best suits the nutrition profile of a given user. A powerful tool to use in food recommendation systems is neural networks. However, the user’s available data are often limited, which compromises the performance of neural-based food recommendation models. To enhance user trust in food recommendations, this paper proposes a method using a secondary model to predict the errors of the primary neural network, especially when dealing with limited data.